MANA - Multi scale adaptive normalized averaging

It is possible to correct intensity inhomogeneity in fat-water Magnetic Resonance Imaging (MRI) by estimating a bias field based on the observed intensities of voxels classified as the pure adipose tissue [1]. The same procedure can also be used to quantify fat volume and its distribution [2] which opens up for new medical applications. The bias field estimation method has to be robust since pure fat voxels are irregularly located and the density varies greatly within and between image volumes. This paper introduces Multi scale Adaptive Normalized Average (MANA) that solves this problem by basing the estimate on a scale space of weighted averages. By using the local certainty of the data MANA preserves details where the local data certainty is high and provides realistic values in sparse areas.

[1]  Magnus Borga,et al.  Intensity Inhomogeneity Correction in Two Point Dixon Imaging , 2008 .

[2]  C. Westin,et al.  Normalized and differential convolution , 1993, Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.

[3]  Klamer Schutte,et al.  Robust Fusion of Irregularly Sampled Data Using Adaptive Normalized Convolution , 2006, EURASIP J. Adv. Signal Process..

[4]  Lucas J. van Vliet,et al.  Normalized Averaging Using Adaptive Applicability Functions with Applications in Image Reconstruction from Sparsely and Randomly Sampled Data , 2003, SCIA.

[5]  Carl-Fredrik Westin,et al.  Normalized and differential convolution , 1993, Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.

[6]  W. T. Dixon.  Simple proton spectroscopic imaging. , 1984, Radiology.

[7]  Örjan Smedby,et al.  Quantitative abdominal fat estimation using MRI , 2008, 2008 19th International Conference on Pattern Recognition.

[8]  Hans Knutsson,et al.  Phase Sensitive Reconstruction for Water/Fat Separation in MR Imaging Using Inverse Gradient , 2007, MICCAI.